Food Cost Variance: Why Your Number Is Probably Wrong (And How to Find the Real Cause)

Read the short version below, then copy the prompt into your company-approved AI tool; start with one site and one stock period. For those looking for the theory behind the prompt and a detailed explanation of why your food cost is probably wrong, keep reading on!

This prompt is built to do one thing first: find the handful of items in your operation that give you the best return on the time you spend tracking them. It won't ask you to count the lettuce. Start with the records you have. The prompt will help you choose what to check first and ask for anything missing. Treat its suggestions as leads to investigate with your team.

After investigating variances across UK hospitality groups, across more than 10,000 stocktakes and millions of supplier orders processed on our platform, the usual suspects of food cost variance (portions, waste or theft) are often innocent.

Before a single portion was over-scooped or before anything hit the bin unlogged, a meaningful share of "bad" variances were already wrong. Usually because a delivery was never processed, an order was placed outside the system or a product was never linked to stock.

This guide is written for UK multi-site operators: the groups big enough for variance to cost real money but small enough that you don't have a full-time analyst chasing it. It covers what variance is, how to check whether yours is real, the seven causes we see most often, a worked investigation from start to finish, and what it actually takes to keep it fixed.

What Is Food Cost Variance?

Food cost variance is the difference between two numbers:

• Theoretical food cost: what you should have spent. Every dish sold in the period, multiplied by its current recipe cost, summed. This is the number your operation would hit with perfect portioning, current prices and zero waste (don't worry, almost no one is!).
• Actual food cost: what you really spent:

Divide each by revenue to get percentages, subtract theoretical from actual, and the difference is your variance. Need the actual figure first? Our guide to how to calculate food cost walks through the formula and the data checks behind it.

What is a good food cost percentage for a UK restaurant? Most full-service UK restaurants aim for 28–32% on food, according to Toast, and the range varies by format. Bar GP varies widely: tied wet-led pubs often run near 50%, while free-of-tie and restaurant bars typically target 70% or more. But the benchmark that matters IS your own best-performing site. If one site runs a 1.5-point variance and its sister site on the same menu runs 5, you don't need an industry table to know where to look.

What is an acceptable food cost variance? Under 2 percentage points: normal tolerance. 2–3 points: monitor the trend. Above 3 points consistently: something systematic is wrong. And note that swinging variance (1% one period, 6% the next) is worse news than steady variance.

Let's attach some real numbers to this to drive the point home. A 3-point variance on a site doing £40,000 a month in food revenue is roughly £1,200 a month, or over £14,000 a year, per site. Across a five-site group, that's a head chef's salary leaking every single year.

The Variance Data Check

Here is the step that separates groups who solve variance from groups who argue about it.

When an operator has a variance that doesn't make sense, we recommend starting by checking whether the data underneath the number is complete; because if it isn't, the variance isn't measuring an operational problem at all, and every hour spent portion-testing the kitchen is an hour wasted investigating a number that ultimately isn't real.

Here are the four checks we recommend running:

Only when all four of these checks come back clean do you have a real variance worth investigating. You can run this check from any system, or from a spreadsheet, in under an hour.

The data check is also why the standard advice fails so many operators. Portion control and waste logging are real levers, but pulling them against a phantom variance changes nothing.

The Seven Common Causes of Food Cost Variance

1. Deliveries received but never processed

The lorry arrives mid-service, the goods go in the fridge, the delivery note goes in a pocket and the order never gets processed.

It rarely gets a mention in variance guides, but it's one of the first things we check. The fix is a receiving discipline: orders are processed at the point of delivery, same day at worst and no stock period closes with arrived-but-unprocessed orders outstanding.

2. Purchases made outside the system

Maybe the head chef rings the mushroom supplier directly, or the bar tops up at Booker. The invoice surfaces weeks, sometimes months, later in a period it doesn't belong to.

One finance director we work with, who runs F&B purchasing across multiple European hotels, watched his sites go nearly £10,000 over on a single month's food cost this way.

GP always looks brilliant when you're not counting all your costs. The fix is absolute: 100% of orders through the system.

3. Portioning above spec

This is typical and one of the most common causes for food cost variance. The tell-tale signature: theoretical says 6kg of cheese should have gone out this period; the count says 10kg went. The reality is every chef running heavy without knowing it, one generous handful at a time, compounding across hundreds of covers.

The fix is scoops, scales at the pass and portion checks as a habit.

4. Unrecorded waste

Waste that isn't logged vanishes from your data and resurfaces as unexplained variance. This includes your mid-week write-offs, the staff meals nobody rang through and the prep trim your recipe yields don't account for.

Quick diagnostic: across 1,130 completed stocktakes on Hops in a single month, exactly half had a recorded waste total of zero. A zero needs a follow-up question: did nothing get thrown away, was it recorded somewhere else, or did nobody write it down? An empty waste log at a busy site is usually the sign of a missing habit.

5. Delivery discrepancies you paid for anyway

If nobody checks the delivery against the purchase order at the back door, your site may end up paying for stock it never received.

The supplier's error is visible on the delivery note, at the moment of delivery. You need someone at the back door checking quantities against the PO before they sign, and raising the credit the same day. And raising it isn't the end: until the supplier accepts it, the credit is pending, and somebody needs to own chasing it (more on this in Keeping It Fixed, below).

6. Supplier price rises nobody caught

Sometimes suppliers move the rates of their produce without notifying operations. This results in recipes not being updated and nobody comparing the invoice to the agreed rate. If this happens, actual has been running above theoretical ever since, and the kitchen, which changed nothing, gets the blame.

The reality is, prices on fresh produce can move weekly. The moment to catch a change is at delivery ("there's a 30% variation on this line and you didn't call us") not at month-end when the P&L lands and the trail is cold.

7. Recipe cards that don't match reality

Here's a scenario you might recognise: the menu changed in September, but the recipe cards didn't, so three sites are still building the dish to the old spec at the old prices. What feels like a kitchen mistake turns out to be a recipe-data problem.

A Worked Investigation: Site B's 4.9-Point Variance

At Hops we like to talk in operator language. Let's walk through a fictitious example of a worked investigation into food cost variance.

The setup: a two-site UK group, casual dining, £38,000/month food revenue per site.

• Site A reports a 1.8-point variance.
• Site B (same menu, same suppliers) reports 4.9.

That's roughly £1,850 of unexplained monthly cost at Site B.

• Total investigation time: roughly five hours.
• Kitchen behaviour changes required: one (portion discipline on one station).
• Number of causes that would've been found by giving a lecture on waste: zero.

Money saved by Steps One and Three: none. They corrected the measurement. The portion fix is the only step that actually reduces cost, and you'll only know by how much when next period's numbers land.

That's the pattern in real life in a real operation. A real variance is usually two or three small causes stacked, and the order you check them in determines whether you find them in five hours or five weeks.

That's the Textbook Answer. At Hops We Deal with Real Life.

Everything above is how variance should be understood, and in a perfect world you'd apply it to every item in every site, every week.

Except you don't live in a perfect world. You run a restaurant group, and time isn't an infinite resource.

The 80/20 of your kitchen

Look at what you buy, not just what's on one plate. In most kitchens a small handful of lines, usually proteins, make up a large share of the spend.

Now imagine it takes the same time and effort to track lettuce properly as it does to track your burger patties. Lettuce might be 1% of your spend. Go from complete chaos to perfect control and the most you can ever recover is that 1%. Get the patties under control and you're working on one of your biggest lines.

That's Pareto's logic in a kitchen: a small number of items carry most of your cost. It doesn't automatically follow that they carry most of your variance. A high-spend line can be tightly controlled and a cheap one can leak every week. So start with purchase value, then look at where the unexplained discrepancy actually sits, in pounds. In the worked investigation it landed in proteins, as it often does, but your list should come from your numbers, not from an assumption.

It's why plenty of experienced operators only track their proteins properly, or why wet-led businesses only count their spirits.

Why fewer items means better control, not worse

This isn't cutting corners: doing any item well takes real discipline, and that discipline is human.

Ask your team to log waste on everything and they'll start dragging their heels within a month. Ask them to log waste on the proteins and they're far more likely to stick with it.

Can't AI just track everything for you? Not really. The analysis can be automated, but the numbers it runs on (the scoop, the weigh, the count, the waste entry) still come from people on shift. Rubbish in, rubbish out.

That's exactly what the prompt at the top is built to do: uncover which items give you the best return on the time you spend tracking them, before it asks you to track anything.

Keeping It Fixed: The Weekly Rhythm

If you know us at Hops, you'll know we bang on about SOPs. Finding the cause once is great, but the key to controlling your food cost variance is all about rhythm and consistency:

Count what's expensive, often

Weekly counts on your 15–20 highest-value lines (proteins, spirits) catch problems in week one instead of month one. Full counts can stay monthly; run wet and dry on whatever cadence you see fit. The weekly counts are checks inside an open period; the close happens when the full count runs.

Close the period properly

A stock period is a fixed window between two counts. "Closing" means every input sits inside that window, and then you sign it off:

Then close it. Everything above exists to get you as close to the truth as possible, so that what's left, the unexplained variance, is as close to zero as possible. Great operators keep it within 1–2 points. Don't be disheartened if it takes a few months and a good number of stocktakes to get there; you're learning the SOPs at the same time as your team. Expect it to improve month on month as you uncover where the variance lives and start explaining it.

And a reminder: anything that surfaces after the close goes into the next period. So you may spend a week scratching your head over a negative variance, then find the case of wine next period and swing positive. Across the two, you're flat. A variance you can explain is fine. It's perfectly normal, in fact; the discipline is in the explaining, not in hitting zero.

Write the exceptions down, not just the routine

Your everyday routines need to run on autopilot: deliveries checked, stock movements recorded, everyone knowing their part. But hospitality produces exceptions every day, and they need equally clear instructions. The supplier sends a substitute. The goods arrive without an invoice. A shortage is recorded but the credit hasn't been accepted. The kitchen makes tomorrow's sandwiches today, an hour before the stocktake.

Each needs an agreed way of being handled:

• Substitutions: record what actually arrived, including the pack size, quantity and price, and link it to the right stock item.
• Missing invoices: record the goods received. Accounts needs an agreed way to recognise their cost in the correct period and reconcile the invoice later.
• Pending credits: keep the claim visible, assign someone to chase it and agree with accounts how it is treated until resolved.
• Food prepared for tomorrow: include the prepared food in the stocktake under the agreed valuation method. Avoid counting both the sandwiches and the ingredients already used to make them.

Those instructions need to be available where the work happens, with unresolved exceptions visible when the period closes. Most "unexplained" variance is an exception that had no rule.

Show the number to the people who control it

Compare sites ruthlessly

In a group, your best site is your benchmark. Same menu, same suppliers, different variance = a process gap, and the fix usually just requires some attention. One group we work with traced a persistent single-site variance to exactly one thing: the person receiving deliveries at that site wasn't checking them. Right process, wrong critical control point.

None of this is super sexy or glamorous, but it works; because it is the fundamental operating system of your restaurant group. It's essentially the sequence Crust Bros ran when they brought discipline to receiving, counting and recipes across their sites, and added around 3% to blended GP as a result.

HOPS is a hospitality operations platform built by operators, for UK restaurant and hotel groups. Book a demo today